diff options
46 files changed, 1488 insertions, 478 deletions
@@ -1,4 +1,4 @@ -Libpng 1.6.28 - January 5, 2017 +Libpng 1.6.29 - March 16, 2017 This is a public release of libpng, intended for use in production codes. @@ -7,28 +7,37 @@ Files available for download: Source files with LF line endings (for Unix/Linux) and with a "configure" script - libpng-1.6.28.tar.xz (LZMA-compressed, recommended) - libpng-1.6.28.tar.gz + libpng-1.6.29.tar.xz (LZMA-compressed, recommended) + libpng-1.6.29.tar.gz Source files with CRLF line endings (for Windows), without the "configure" script - lpng1628.7z (LZMA-compressed, recommended) - lpng1628.zip + lpng1629.7z (LZMA-compressed, recommended) + lpng1629.zip Other information: - libpng-1.6.28-README.txt - libpng-1.6.28-LICENSE.txt - libpng-1.6.28-*.asc (armored detached GPG signatures) - -Changes since the last public release (1.6.27): - Fixed arm/aarch64 detection in CMakeLists.txt (Gianfranco Costamagna). - Added option to Cmake build allowing a custom location of zlib to be - specified in a scenario where libpng is being built as a subproject - alongside zlib by another project (Sam Serrels). - Changed png_ptr->options from a png_byte to png_uint_32, to accomodate - up to 16 options. + libpng-1.6.29-README.txt + libpng-1.6.29-LICENSE.txt + libpng-1.6.29-*.asc (armored detached GPG signatures) + +Changes since the last public release (1.6.28): + + Readded "include(GNUInstallDirs)" to CMakeLists.txt (Gianfranco Costamagna). + Moved SSE2 optimization code into the main libpng source directory. + Configure libpng with "configure --enable-intel-sse" or compile + libpng with "-DPNG_INTEL_SSE" in CPPFLAGS to enable it. + Simplified conditional compilation in pngvalid.c, for AIX (Michael Felt). + Avoid conditional directives that break statements in pngrutil.c (Romero + Malaquias) + The contrib/examples/pngtopng.c recovery code was in the wrong "if" + branches; the comments were correct. + Added code for PowerPC VSX optimisation (Vadim Barkov). + Avoid potential overflow of shift operations in png_do_expand() (Aaron Boxer). + Change test ZLIB_VERNUM >= 0x1281 to ZLIB_VERNUM >= 0x1290 in pngrutil.c + because Solaris 11 distributes zlib-1.2.8.f that is older than 1.2.8.1. + Suppress clang warnings about implicit sign changes in png.c Send comments/corrections/commendations to png-mng-implement at lists.sf.net (subscription required; visit @@ -5794,6 +5794,29 @@ Version 1.6.28rc03 [January 4, 2017] Version 1.6.28 [January 5, 2017] No changes. +Version 1.6.29beta01 [January 12, 2017] + Readded "include(GNUInstallDirs)" to CMakeLists.txt (Gianfranco Costamagna). + Moved SSE2 optimization code into the main libpng source directory. + Configure libpng with "configure --enable-intel-sse" or compile + libpng with "-DPNG_INTEL_SSE" in CPPFLAGS to enable it. + Simplified conditional compilation in pngvalid.c, for AIX (Michael Felt). + +Version 1.6.29beta02 [February 22, 2017] + Avoid conditional directives that break statements in pngrutil.c (Romero + Malaquias) + The contrib/examples/pngtopng.c recovery code was in the wrong "if" + branches; the comments were correct. + Added code for PowerPC VSX optimisation (Vadim Barkov). + +Version 1.6.29beta03 [March 1, 2017] + Avoid potential overflow of shift operations in png_do_expand() (Aaron Boxer). + Change test ZLIB_VERNUM >= 0x1281 to ZLIB_VERNUM >= 0x1290 in pngrutil.c + because Solaris 11 distributes zlib-1.2.8.f that is older than 1.2.8.1. + Suppress clang warnings about implicit sign changes in png.c + +Version 1.6.29 [March 16, 2017] + No changes. + Send comments/corrections/commendations to png-mng-implement at lists.sf.net (subscription required; visit https://lists.sourceforge.net/lists/listinfo/png-mng-implement diff --git a/CMakeLists.txt b/CMakeLists.txt index e076d5e94..3b6c1f30d 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -36,13 +36,15 @@ enable_testing() set(PNGLIB_MAJOR 1) set(PNGLIB_MINOR 6) -set(PNGLIB_RELEASE 28) +set(PNGLIB_RELEASE 29) set(PNGLIB_NAME libpng${PNGLIB_MAJOR}${PNGLIB_MINOR}) set(PNGLIB_VERSION ${PNGLIB_MAJOR}.${PNGLIB_MINOR}.${PNGLIB_RELEASE}) +include(GNUInstallDirs) + # needed packages -#Allow users to specify location of Zlib, +#Allow users to specify location of Zlib, # Useful if zlib is being built alongside this as a sub-project option(PNG_BUILD_ZLIB "Custom zlib Location, else find_package is used" OFF) @@ -96,7 +98,7 @@ if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^arm" OR arm/arm_init.c arm/filter_neon.S arm/filter_neon_intrinsics.c) - + if(${PNG_ARM_NEON} STREQUAL "on") add_definitions(-DPNG_ARM_NEON_OPT=2) elseif(${PNG_ARM_NEON} STREQUAL "check") @@ -107,6 +109,38 @@ if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^arm" OR endif() endif() +# set definitions and sources for powerpc +if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^powerpc*" OR + ${CMAKE_SYSTEM_PROCESSOR} MATCHES "^ppc64*" ) + set(PNG_POWERPC_VSX_POSSIBLE_VALUES check on off) + set(PNG_POWERPC_VSX "check" CACHE STRING "Enable POWERPC VSX optimizations: + check: (default) use internal checking code; + off: disable the optimizations; + on: turn on unconditionally.") + set_property(CACHE PNG_POWERPC_VSX PROPERTY STRINGS + ${PNG_POWERPC_VSX_POSSIBLE_VALUES}) + list(FIND PNG_POWERPC_VSX_POSSIBLE_VALUES ${PNG_POWERPC_VSX} index) + if(index EQUAL -1) + message(FATAL_ERROR + " PNG_POWERPC_VSX must be one of [${PNG_POWERPC_VSX_POSSIBLE_VALUES}]") + elseif(NOT ${PNG_POWERPC_VSX} STREQUAL "no") + set(libpng_powerpc_sources + powerpc/powerpc_init.c + powerpc/filter_vsx_intrinsics.c) + + if(${PNG_POWERPC_VSX} STREQUAL "on") + add_definitions(-DPNG_POWERPC_VSX_OPT=2) + elseif(${PNG_POWERPC_VSX} STREQUAL "check") + add_definitions(-DPNG_POWERPC_VSX_CHECK_SUPPORTED) + message(WARNING + "[PNG_POWERPC_VSX==check] Please check contrib/powerpc/README file for the list of supported OSes.") + endif() + else() + add_definitions(-DPNG_POWERPC_VSX_OPT=0) + endif() +endif() + + # SET LIBNAME set(PNG_LIB_NAME png${PNGLIB_MAJOR}${PNGLIB_MINOR}) @@ -398,6 +432,7 @@ set(libpng_sources pngwtran.c pngwutil.c ${libpng_arm_sources} + ${libpng_powerpc_sources} ) set(pngtest_sources pngtest.c @@ -735,7 +770,7 @@ endif(NOT WIN32 OR CYGWIN OR MINGW) # SET UP LINKS if(PNG_SHARED) set_target_properties(png PROPERTIES -# VERSION 16.${PNGLIB_RELEASE}.1.6.28 +# VERSION 16.${PNGLIB_RELEASE}.1.6.29 VERSION 16.${PNGLIB_RELEASE}.0 SOVERSION 16 CLEAN_DIRECT_OUTPUT 1) @@ -840,4 +875,3 @@ endif() # to create msvc import lib for mingw compiled shared lib # pexports libpng.dll > libpng.def # lib /def:libpng.def /machine:x86 - @@ -10,7 +10,7 @@ this sentence. This code is released under the libpng license. -libpng versions 1.0.7, July 1, 2000 through 1.6.28, January 5, 2017 are +libpng versions 1.0.7, July 1, 2000 through 1.6.29, March 16, 2017 are Copyright (c) 2000-2002, 2004, 2006-2017 Glenn Randers-Pehrson, are derived from libpng-1.0.6, and are distributed according to the same disclaimer and license as libpng-1.0.6 with the following individuals @@ -23,6 +23,8 @@ added to the list of Contributing Authors: Gilles Vollant James Yu Mandar Sahastrabuddhe + Google Inc. + Vadim Barkov and with the following additions to the disclaimer: @@ -128,4 +130,4 @@ any encryption software. See the EAR, paragraphs 734.3(b)(3) and Glenn Randers-Pehrson glennrp at users.sourceforge.net -January 5, 2017 +March 16, 2017 diff --git a/Makefile.am b/Makefile.am index fb209edd2..18ac74866 100644 --- a/Makefile.am +++ b/Makefile.am @@ -102,6 +102,16 @@ libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += mips/mips_init.c\ mips/filter_msa_intrinsics.c endif +if PNG_INTEL_SSE +libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += intel/intel_init.c\ + intel/filter_sse2_intrinsics.c +endif + +if PNG_POWERPC_VSX +libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += powerpc/powerpc_init.c\ + powerpc/filter_vsx_intrinsics.c +endif + nodist_libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES = pnglibconf.h libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_LDFLAGS = -no-undefined -export-dynamic \ @@ -1,4 +1,4 @@ -README for libpng version 1.6.28 - January 5, 2017 (shared library 16.0) +README for libpng version 1.6.29 - March 16, 2017 (shared library 16.0) See the note about version numbers near the top of png.h See INSTALL for instructions on how to install libpng. @@ -179,14 +179,16 @@ Files in this distribution: pngwtran.c => Write data transformations pngwutil.c => Write utility functions arm => Contains optimized code for the ARM platform + powerpc => Contains optimized code for the PowerPC platform contrib => Contributions arm-neon => Optimized code for ARM-NEON platform + powerpc-vsx => Optimized code for POWERPC-VSX platform examples => Example programs gregbook => source code for PNG reading and writing, from Greg Roelofs' "PNG: The Definitive Guide", O'Reilly, 1999 - intel => Optimized code for INTEL-SSE2 platform libtests => Test programs + mips-msa => Optimized code for MIPS-MSA platform pngminim => Minimal decoder, encoder, and progressive decoder programs demonstrating use of pngusr.dfa pngminus => Simple pnm2png and png2pnm programs @@ -194,6 +196,8 @@ Files in this distribution: testpngs tools => Various tools visupng => Contains a MSVC workspace for VisualPng + intel => Optimized code for INTEL-SSE2 platform + mips => Optimized code for MIPS platform projects => Contains project files and workspaces for building a DLL owatcom => Contains a WATCOM project for building libpng @@ -25,5 +25,6 @@ Use greater precision when changing to linear gamma for compositing against Investigate pre-incremented loop counters and other loop constructions. Add interpolated method of handling interlacing. Extend pngvalid.c to validate more of the libpng transformations. +Refactor preprocessor conditionals to compile entire statements */ diff --git a/configure.ac b/configure.ac index 8c474a6d7..765a4ad2d 100644 --- a/configure.ac +++ b/configure.ac @@ -25,7 +25,7 @@ AC_PREREQ([2.68]) dnl Version number stuff here: -AC_INIT([libpng],[1.6.28],[png-mng-implement@lists.sourceforge.net]) +AC_INIT([libpng],[1.6.29],[png-mng-implement@lists.sourceforge.net]) AC_CONFIG_MACRO_DIR([scripts]) # libpng does not follow GNU file name conventions (hence 'foreign') @@ -46,7 +46,7 @@ dnl automake, so the following is not necessary (and is not defined anyway): dnl AM_PREREQ([1.11.2]) dnl stop configure from automagically running automake -PNGLIB_VERSION=1.6.28 +PNGLIB_VERSION=1.6.29 PNGLIB_MAJOR=1 PNGLIB_MINOR=6 PNGLIB_RELEASE=28 @@ -391,6 +391,89 @@ AM_CONDITIONAL([PNG_MIPS_MSA], mipsel*|mips64el*) :;; esac]) +# INTEL +# ===== +# +# INTEL SSE (SIMD) support. + +AC_ARG_ENABLE([intel-sse], + AS_HELP_STRING([[[--enable-intel-sse]]], + [Enable Intel SSE optimizations: =no/off, yes/on:] + [no/off: disable the optimizations;] + [yes/on: enable the optimizations.] + [If not specified: determined by the compiler.]), + [case "$enableval" in + no|off) + # disable the default enabling: + AC_DEFINE([PNG_INTEL_SSE_OPT], [0], + [Disable Intel SSE optimizations]) + # Prevent inclusion of the assembler files below: + enable_intel_sse=no;; + yes|on) + AC_DEFINE([PNG_INTEL_SSE_OPT], [1], + [Enable Intel SSE optimizations]);; + *) + AC_MSG_ERROR([--enable-intel-sse=${enable_intel_sse}: invalid value]) + esac]) + +# Add Intel specific files to all builds where the host_cpu is Intel ('x86*') +# or where Intel optimizations were explicitly requested (this allows a +# fallback if a future host CPU does not match 'x86*') +AM_CONDITIONAL([PNG_INTEL_SSE], + [test "$enable_intel_sse" != 'no' && + case "$host_cpu" in + i?86|x86_64) :;; + *) test "$enable_intel_sse" != '';; + esac]) + +# PowerPC +# === +# +# PowerPC VSX (SIMD) support. + +AC_ARG_ENABLE([powerpc-vsx], +AS_HELP_STRING([[[--enable-powerpc-vsx]]], + [Enable POWERPC VSX optimizations: =no/off, check, api, yes/on:] + [no/off: disable the optimizations; check: use internal checking code] + [api: disable by default, enable by a call to png_set_option] + [yes/on: turn on unconditionally.] + [If not specified: determined by the compiler.]), + [case "$enableval" in + no|off) + # disable the default enabling on __ppc64__ systems: + AC_DEFINE([PNG_POWERPC_VSX_OPT], [0], + [Disable POWERPC VSX optimizations]) + # Prevent inclusion of the platform specific files below: + enable_powerpc_vsx=no;; + check) + AC_DEFINE([PNG_POWERPC_VSX_CHECK_SUPPORTED], [], + [Check for POWERPC VSX support at run-time]) + AC_MSG_WARN([--enable-powerpc-vsx Please check contrib/powerpc/README file] + [for the list of supported OSes.]);; + api) + AC_DEFINE([PNG_POWERPC_VSX_API_SUPPORTED], [], + [Turn on POWERPC VSX optimizations at run-time]);; + yes|on) + AC_DEFINE([PNG_POWERPC_VSX_OPT], [2], + [Enable POWERPC VSX optimizations]) + AC_MSG_WARN([--enable-powerpc-vsx: please specify 'check' or 'api', if] + [you want the optimizations unconditionally pass '-maltivec -mvsx'] + [or '-mcpu=power8'to the compiler.]);; + *) + AC_MSG_ERROR([--enable-powerpc-vsx=${enable_powerpc_vsx}: invalid value]) + esac]) + +# Add PowerPC specific files to all builds where the host_cpu is powerpc('powerpc*') or +# where POWERPC optimizations were explicitly requested (this allows a fallback if a +# future host CPU does not match 'powerpc*') + +AM_CONDITIONAL([PNG_POWERPC_VSX], + [test "$enable_powerpc_vsx" != 'no' && + case "$host_cpu" in + powerpc*|ppc64*) :;; + esac]) + + AC_MSG_NOTICE([[Extra options for compiler: $PNG_COPTS]]) # Config files, substituting as above diff --git a/contrib/examples/pngtopng.c b/contrib/examples/pngtopng.c index 4acf6b3ad..d4a51a8f0 100644 --- a/contrib/examples/pngtopng.c +++ b/contrib/examples/pngtopng.c @@ -1,10 +1,12 @@ /*- pngtopng * - * COPYRIGHT: Written by John Cunningham Bowler, 2011. + * COPYRIGHT: Written by John Cunningham Bowler, 2011, 2017. * To the extent possible under law, the author has waived all copyright and * related or neighboring rights to this work. This work is published from: * United States. * + * Last changed in libpng 1.6.29 [%RDATE%] + * * Read a PNG and write it out in a fixed format, using the 'simplified API' * that was introduced in libpng-1.6.0. * @@ -59,26 +61,27 @@ int main(int argc, const char **argv) else fprintf(stderr, "pngtopng: write %s: %s\n", argv[2], image.message); - - free(buffer); } else - { fprintf(stderr, "pngtopng: read %s: %s\n", argv[1], image.message); - /* This is the only place where a 'free' is required; libpng does - * the cleanup on error and success, but in this case we couldn't - * complete the read because of running out of memory. - */ - png_image_free(&image); - } + free(buffer); } else + { fprintf(stderr, "pngtopng: out of memory: %lu bytes\n", (unsigned long)PNG_IMAGE_SIZE(image)); + + /* This is the only place where a 'free' is required; libpng does + * the cleanup on error and success, but in this case we couldn't + * complete the read because of running out of memory and so libpng + * has not got to the point where it can do cleanup. + */ + png_image_free(&image); + } } else diff --git a/contrib/intel/INSTALL b/contrib/intel/INSTALL deleted file mode 100644 index cd5cdd94e..000000000 --- a/contrib/intel/INSTALL +++ /dev/null @@ -1,158 +0,0 @@ -Enabling SSE support - -Copyright (c) 2016 Google, Inc. -Written by Mike Klein, Matt Sarett - -This INSTALL file written by Glenn Randers-Pehrson, 2016. - -If you have moved intel_init.c and filter_sse2_intrinsics.c to a different -directory, be sure to update the '#include "../../pngpriv.h"' line in both -files if necessary to point to the correct relative location of pngpriv.h -with respect to the new location of those files. - -To enable SSE support in libpng, follow the instructions in I, II, or III, -below: - -I. Using patched "configure" scripts: - -First, apply intel_sse.patch in your build directory. - - patch -i contrib/intel/intel_sse.patch -p1 - -Then, if you are not building in a new GIT clone, e.g., in a tar -distribution, remove any existing pre-built configure scripts: - - ./configure --enable-maintainer-mode - make maintainer-clean - ./autogen.sh --maintainer --clean - -Finally, configure libpng with -DPNG_INTEL_SSE in CPPFLAGS: - - ./autogen.sh --maintainer - CPPFLAGS="-DPNG_INTEL_SSE" ./configure [options] - make CPPFLAGS="-DPNG_INTEL_SSE" [options] - make - -II. Using a custom makefile: - -If you are using a custom makefile makefile, you will have to update it -manually to include contrib/intel/*.o in the dependencies, and to define -PNG_INTEL_SSE. - -III. Using manually updated "configure" scripts: - -If you prefer, manually edit pngpriv.h, configure.ac, and Makefile.am, -following the instructions below, then follow the instructions in -section II of INSTALL in the main libpng directory, then configure libpng -with -DPNG_INTEL_SSE in CPPFLAGS. - -1. Add the following code to configure.ac under HOST SPECIFIC OPTIONS -directly beneath the section for ARM: - ------------------cut---------------- -# INTEL -# ===== -# -# INTEL SSE (SIMD) support. - -AC_ARG_ENABLE([intel-sse], - AS_HELP_STRING([[[--enable-intel-sse]]], - [Enable Intel SSE optimizations: =no/off, yes/on:] - [no/off: disable the optimizations;] - [yes/on: enable the optimizations.] - [If not specified: determined by the compiler.]), - [case "$enableval" in - no|off) - # disable the default enabling: - AC_DEFINE([PNG_INTEL_SSE_OPT], [0], - [Disable Intel SSE optimizations]) - # Prevent inclusion of the assembler files below: - enable_intel_sse=no;; - yes|on) - AC_DEFINE([PNG_INTEL_SSE_OPT], [1], - [Enable Intel SSE optimizations]);; - *) - AC_MSG_ERROR([--enable-intel-sse=${enable_intel_sse}: invalid value]) - esac]) - -# Add Intel specific files to all builds where the host_cpu is Intel ('x86*') -# or where Intel optimizations were explicitly requested (this allows a -# fallback if a future host CPU does not match 'x86*') -AM_CONDITIONAL([PNG_INTEL_SSE], - [test "$enable_intel_sse" != 'no' && - case "$host_cpu" in - i?86|x86_64) :;; - *) test "$enable_intel_sse" != '';; - esac]) ------------------cut---------------- - -2. Add the following code to Makefile.am under HOST SPECIFIC OPTIONS -directly beneath the "if PNG_ARM_NEON ... endif" statement: - ------------------cut---------------- -if PNG_INTEL_SSE -libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += contrib/intel/intel_init.c\ - contrib/intel/filter_sse2_intrinsics.c -endif ------------------cut---------------- - -3. Add the following lines to pngpriv.h, following the PNG_ARM_NEON_OPT -code: - ------------------cut---------------- -#ifndef PNG_INTEL_SSE_OPT -# ifdef PNG_INTEL_SSE - /* Only check for SSE if the build configuration has been modified to - * enable SSE optimizations. This means that these optimizations will - * be off by default. See contrib/intel for more details. - */ -# if defined(__SSE4_1__) || defined(__AVX__) || defined(__SSSE3__) || \ - defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \ - (defined(_M_IX86_FP) && _M_IX86_FP >= 2) -# define PNG_INTEL_SSE_OPT 1 -# endif -# endif -#endif - -#if PNG_INTEL_SSE_OPT > 0 -# ifndef PNG_INTEL_SSE_IMPLEMENTATION -# if defined(__SSE4_1__) || defined(__AVX__) - /* We are not actually using AVX, but checking for AVX is the best - way we can detect SSE4.1 and SSSE3 on MSVC. - */ -# define PNG_INTEL_SSE_IMPLEMENTATION 3 -# elif defined(__SSSE3__) -# define PNG_INTEL_SSE_IMPLEMENTATION 2 -# elif defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \ - (defined(_M_IX86_FP) && _M_IX86_FP >= 2) -# define PNG_INTEL_SSE_IMPLEMENTATION 1 -# else -# define PNG_INTEL_SSE_IMPLEMENTATION 0 -# endif -# endif - -# if PNG_INTEL_SSE_IMPLEMENTATION > 0 -# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_sse2 -# endif -#endif - ------------------cut---------------- - -4. Add the following lines to pngpriv.h, following the prototype for -png_read_filter_row_paeth4_neon: - ------------------cut---------------- -PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_sse2,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_sse2,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_sse2,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_sse2,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_sse2,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_sse2,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); - ------------------cut---------------- diff --git a/contrib/intel/intel_sse.patch b/contrib/intel/intel_sse.patch deleted file mode 100644 index 24d4883c7..000000000 --- a/contrib/intel/intel_sse.patch +++ /dev/null @@ -1,190 +0,0 @@ -diff --git a/configure.ac b/configure.ac ---- a/configure.ac 2016-08-29 11:46:27.000000000 -0400 -+++ b/configure.ac 2016-08-29 16:57:03.866355018 -0400 -@@ -386,16 +386,51 @@ AC_ARG_ENABLE([mips-msa], - # future host CPU does not match 'mips*') - - AM_CONDITIONAL([PNG_MIPS_MSA], - [test "$enable_mips_msa" != 'no' && - case "$host_cpu" in - mipsel*|mips64el*) :;; - esac]) - -+# INTEL -+# ===== -+# -+# INTEL SSE (SIMD) support. -+ -+AC_ARG_ENABLE([intel-sse], -+ AS_HELP_STRING([[[--enable-intel-sse]]], -+ [Enable Intel SSE optimizations: =no/off, yes/on:] -+ [no/off: disable the optimizations;] -+ [yes/on: enable the optimizations.] -+ [If not specified: determined by the compiler.]), -+ [case "$enableval" in -+ no|off) -+ # disable the default enabling: -+ AC_DEFINE([PNG_INTEL_SSE_OPT], [0], -+ [Disable Intel SSE optimizations]) -+ # Prevent inclusion of the assembler files below: -+ enable_intel_sse=no;; -+ yes|on) -+ AC_DEFINE([PNG_INTEL_SSE_OPT], [1], -+ [Enable Intel SSE optimizations]);; -+ *) -+ AC_MSG_ERROR([--enable-intel-sse=${enable_intel_sse}: invalid value]) -+ esac]) -+ -+# Add Intel specific files to all builds where the host_cpu is Intel ('x86*') -+# or where Intel optimizations were explicitly requested (this allows a -+# fallback if a future host CPU does not match 'x86*') -+AM_CONDITIONAL([PNG_INTEL_SSE], -+ [test "$enable_intel_sse" != 'no' && -+ case "$host_cpu" in -+ i?86|x86_64) :;; -+ *) test "$enable_intel_sse" != '';; -+ esac]) -+ - AC_MSG_NOTICE([[Extra options for compiler: $PNG_COPTS]]) - - # Config files, substituting as above - AC_CONFIG_FILES([Makefile libpng.pc:libpng.pc.in]) - AC_CONFIG_FILES([libpng-config:libpng-config.in], - [chmod +x libpng-config]) - - AC_OUTPUT -diff --git a/Makefile.am b/Makefile.am ---- a/Makefile.am 2016-08-29 11:46:27.000000000 -0400 -+++ b/Makefile.am 2016-08-29 16:57:45.955528215 -0400 -@@ -97,16 +97,21 @@ libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SO - arm/filter_neon.S arm/filter_neon_intrinsics.c - endif - - if PNG_MIPS_MSA - libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += mips/mips_init.c\ - mips/filter_msa_intrinsics.c - endif - -+if PNG_INTEL_SSE -+libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES += contrib/intel/intel_init.c\ -+ contrib/intel/filter_sse2_intrinsics.c -+endif -+ - nodist_libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_SOURCES = pnglibconf.h - - libpng@PNGLIB_MAJOR@@PNGLIB_MINOR@_la_LDFLAGS = -no-undefined -export-dynamic \ - -version-number @PNGLIB_MAJOR@@PNGLIB_MINOR@:@PNGLIB_RELEASE@:0 - - if HAVE_LD_VERSION_SCRIPT - # Versioned symbols and restricted exports - if HAVE_SOLARIS_LD -diff --git a/pngpriv.h b/pngpriv.h ---- debug16/pngpriv.h 2016-08-30 10:46:36.000000000 -0400 -+++ libpng16/pngpriv.h 2016-08-30 11:57:25.672280202 -0400 -@@ -185,16 +185,52 @@ - #ifndef PNG_MIPS_MSA_OPT - # if defined(__mips_msa) && (__mips_isa_rev >= 5) && defined(PNG_ALIGNED_MEMORY_SUPPORTED) - # define PNG_MIPS_MSA_OPT 2 - # else - # define PNG_MIPS_MSA_OPT 0 - # endif - #endif - -+#ifndef PNG_INTEL_SSE_OPT -+# ifdef PNG_INTEL_SSE -+ /* Only check for SSE if the build configuration has been modified to -+ * enable SSE optimizations. This means that these optimizations will -+ * be off by default. See contrib/intel for more details. -+ */ -+# if defined(__SSE4_1__) || defined(__AVX__) || defined(__SSSE3__) || \ -+ defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \ -+ (defined(_M_IX86_FP) && _M_IX86_FP >= 2) -+# define PNG_INTEL_SSE_OPT 1 -+# endif -+# endif -+#endif -+ -+#if PNG_INTEL_SSE_OPT > 0 -+# ifndef PNG_INTEL_SSE_IMPLEMENTATION -+# if defined(__SSE4_1__) || defined(__AVX__) -+ /* We are not actually using AVX, but checking for AVX is the best -+ way we can detect SSE4.1 and SSSE3 on MSVC. -+ */ -+# define PNG_INTEL_SSE_IMPLEMENTATION 3 -+# elif defined(__SSSE3__) -+# define PNG_INTEL_SSE_IMPLEMENTATION 2 -+# elif defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \ -+ (defined(_M_IX86_FP) && _M_IX86_FP >= 2) -+# define PNG_INTEL_SSE_IMPLEMENTATION 1 -+# else -+# define PNG_INTEL_SSE_IMPLEMENTATION 0 -+# endif -+# endif -+ -+# if PNG_INTEL_SSE_IMPLEMENTATION > 0 -+# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_sse2 -+# endif -+#endif -+ - #if PNG_MIPS_MSA_OPT > 0 - # define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_msa - # ifndef PNG_MIPS_MSA_IMPLEMENTATION - # if defined(__mips_msa) - # if defined(__clang__) - # elif defined(__GNUC__) - # if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 7) - # define PNG_MIPS_MSA_IMPLEMENTATION 2 -@@ -1251,16 +1287,31 @@ PNG_INTERNAL_FUNCTION(void,png_read_filt - PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_msa,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); - PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_msa,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); - PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_msa,(png_row_infop - row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); - #endif - -+#if PNG_INTEL_SSE_IMPLEMENTATION > 0 -+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_sse2,(png_row_infop -+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_sse2,(png_row_infop -+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_sse2,(png_row_infop -+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_sse2,(png_row_infop -+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_sse2,(png_row_infop -+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -+PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_sse2,(png_row_infop -+ row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); -+#endif -+ - /* Choose the best filter to use and filter the row data */ - PNG_INTERNAL_FUNCTION(void,png_write_find_filter,(png_structrp png_ptr, - png_row_infop row_info),PNG_EMPTY); - - #ifdef PNG_SEQUENTIAL_READ_SUPPORTED - PNG_INTERNAL_FUNCTION(void,png_read_IDAT_data,(png_structrp png_ptr, - png_bytep output, png_alloc_size_t avail_out),PNG_EMPTY); - /* Read 'avail_out' bytes of data from the IDAT stream. If the output buffer -@@ -1986,16 +2037,21 @@ PNG_INTERNAL_FUNCTION(void, PNG_FILTER_O - PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_neon, - (png_structp png_ptr, unsigned int bpp), PNG_EMPTY); - #endif - - #if PNG_MIPS_MSA_OPT > 0 - PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_msa, - (png_structp png_ptr, unsigned int bpp), PNG_EMPTY); - #endif -+ -+# if PNG_INTEL_SSE_IMPLEMENTATION > 0 -+PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_sse2, -+ (png_structp png_ptr, unsigned int bpp), PNG_EMPTY); -+# endif - #endif - - PNG_INTERNAL_FUNCTION(png_uint_32, png_check_keyword, (png_structrp png_ptr, - png_const_charp key, png_bytep new_key), PNG_EMPTY); - - /* Maintainer: Put new private prototypes here ^ */ - - #include "pngdebug.h" diff --git a/contrib/libtests/pngvalid.c b/contrib/libtests/pngvalid.c index 499801b30..b0daa18b1 100644 --- a/contrib/libtests/pngvalid.c +++ b/contrib/libtests/pngvalid.c @@ -1,8 +1,8 @@ /* pngvalid.c - validate libpng by constructing then reading png files. * - * Last changed in libpng 1.6.27 [December 29, 2016] - * Copyright (c) 2014-2016 John Cunningham Bowler + * Last changed in libpng 1.6.29 [March 16, 2017] + * Copyright (c) 2014-2017 John Cunningham Bowler * * This code is released under the libpng license. * For conditions of distribution and use, see the disclaimer @@ -7742,13 +7742,11 @@ image_transform_png_set_rgb_to_gray_ini(const image_transform *this, * NOTE: this number only affects the internal limit check in pngvalid, * it has no effect on the limits applied to the libpng values. */ - that->pm->limit += pow( -# if DIGITIZE - 2.0 -# else - 1.0 -# endif - /255, data.gamma); +#if DIGITIZE + that->pm->limit += pow( 2.0/255, data.gamma); +#else + that->pm->limit += pow( 1.0/255, data.gamma); +#endif } } diff --git a/contrib/powerpc-vsx/README b/contrib/powerpc-vsx/README new file mode 100644 index 000000000..e566147ea --- /dev/null +++ b/contrib/powerpc-vsx/README @@ -0,0 +1,81 @@ +OPERATING SYSTEM SPECIFIC POWERPC DETECTION +-------------------------------------------- + +Detection of the ability to execute POWERPC on processor requires +operating system support. (The information is not available in user mode.) + +Currently only this feature is supported only for linux platform. + +HOW TO USE THIS +--------------- + +This directory contains C code fragments that can be included in powerpc/powerpc_init.c +by setting the macro PNG_POWERPC_VSX_FILE to the file name in "" or <> at build +time. This setting is not recorded in pnglibconf.h and can be changed simply by +rebuilding arm/arm_init.o with the required macro definition. + +For any of this code to be used the POWERPC code must be enabled and run time +checks must be supported. I.e.: + +#if PNG_POWERPC_VSX_OPT > 0 +#ifdef PNG_POWERPC_VSX_CHECK_SUPPORTED + +This is done in a 'configure' build by passing configure the argument: + + --enable-powerpc-vsx=check + +FILE FORMAT +----------- + +Each file documents its testing status as of the last time it was tested (which +may have been a long time ago): + +STATUS: one of: + SUPPORTED: This indicates that the file is included in the regularly + performed test builds and bugs are fixed when discovered. + COMPILED: This indicates that the code did compile at least once. See the + more detailed description for the extent to which the result was + successful. + TESTED: This means the code was fully compiled into the libpng test programs + and these were run at least once. + +BUG REPORTS: an email address to which to send reports of problems + +The file is a fragment of C code. It should not define any 'extern' symbols; +everything should be static. It must define the function: + +static int png_have_vsx(png_structp png_ptr); + +That function must return 1 if ARM NEON instructions are supported, 0 if not. +It must not execute png_error unless it detects a bug. A png_error will prevent +the reading of the PNG and in the future, writing too. + +BUG REPORTS +----------- + +If you mail a bug report for any file that is not SUPPORTED there may only be +limited response. Consider fixing it and sending a patch to fix the problem - +this is more likely to result in action. + +CONTRIBUTIONS +------------- + +You may send contributions of new implementations to +png-mng-implement@sourceforge.net. Please write code in strict C90 C where +possible. Obviously OS dependencies are to be expected. If you submit code you +must have the authors permission and it must have a license that is acceptable +to the current maintainer; in particular that license must permit modification +and redistribution. + +Please try to make the contribution a single file and give the file a clear and +unambiguous name that identifies the target OS. If multiple files really are +required put them all in a sub-directory. + +You must also be prepared to handle bug reports from users of the code, either +by joining the png-mng-implement mailing list or by providing an email for the +"BUG REPORTS" entry or both. Please make sure that the header of the file +contains the STATUS and BUG REPORTS fields as above. + +Please list the OS requirements as precisely as possible. Ideally you should +also list the environment in which the code has been tested and certainly list +any environments where you suspect it might not work. diff --git a/contrib/powerpc-vsx/linux.c b/contrib/powerpc-vsx/linux.c new file mode 100644 index 000000000..32ed9d749 --- /dev/null +++ b/contrib/powerpc-vsx/linux.c @@ -0,0 +1,57 @@ +/* contrib/powerpc-vsx/linux.c + * + * Copyright (c) 2017 Glenn Randers-Pehrson + * Written by Vadim Barkov, 2017. + * Last changed in libpng 1.6.29 [March 16, 2017] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * STATUS: TESTED + * BUG REPORTS: png-mng-implement@sourceforge.net + * + * png_have_vsx implemented for Linux by reading the widely available + * pseudo-file /proc/cpuinfo. + * + * This code is strict ANSI-C and is probably moderately portable; it does + * however use <stdio.h> and it assumes that /proc/cpuinfo is never localized. + */ + +#include <stdio.h> +#include <string.h> +#include <stdlib.h> +#include "png.h" + +#ifndef MAXLINE +# define MAXLINE 1024 +#endif + +static int +png_have_vsx(png_structp png_ptr) +{ + FILE *f; + + const char *string = "altivec supported"; + char input[MAXLINE]; + char *token = NULL; + + PNG_UNUSED(png_ptr) + + f = fopen("/proc/cpuinfo", "r"); + if (f != NULL) + { + memset(input,0,MAXLINE); + while(fgets(input,MAXLINE,f) != NULL) + { + token = strstr(input,string); + if(token != NULL) + return 1; + } + } +#ifdef PNG_WARNINGS_SUPPORTED + else + png_warning(png_ptr, "/proc/cpuinfo open failed"); +#endif + return 0; +} diff --git a/contrib/powerpc-vsx/linux_aux.c b/contrib/powerpc-vsx/linux_aux.c new file mode 100644 index 000000000..796922d77 --- /dev/null +++ b/contrib/powerpc-vsx/linux_aux.c @@ -0,0 +1,36 @@ +/* contrib/powerpc-vsx/linux_aux.c + * + * Copyright (c) 2017 Glenn Randers-Pehrson + * Written by Vadim Barkov, 2017. + * Last changed in libpng 1.6.29 [March 16, 2017] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * STATUS: TESTED + * BUG REPORTS: png-mng-implement@sourceforge.net + * + * png_have_vsx implemented for Linux by using the auxiliary vector mechanism. + * + * This code is strict ANSI-C and is probably moderately portable; it does + * however use <stdio.h> and it assumes that /proc/cpuinfo is never localized. + */ + +#include "sys/auxv.h" +#include "png.h" + +static int +png_have_vsx(png_structp png_ptr) +{ + + const unsigned long auxv = getauxval( AT_HWCAP ); + + PNG_UNUSED(png_ptr) + + if(auxv & (PPC_FEATURE_HAS_ALTIVEC|PPC_FEATURE_HAS_VSX )) + return 1; + else + return 0; +} + diff --git a/contrib/intel/filter_sse2_intrinsics.c b/intel/filter_sse2_intrinsics.c index b02840c55..b91d8da96 100644 --- a/contrib/intel/filter_sse2_intrinsics.c +++ b/intel/filter_sse2_intrinsics.c @@ -1,19 +1,18 @@ /* filter_sse2_intrinsics.c - SSE2 optimized filter functions * - * Copyright (c) 2016 Google, Inc. + * Copyright (c) 2016-2017 Glenn Randers-Pehrson * Written by Mike Klein and Matt Sarett - * Derived from arm/filter_neon_intrinsics.c, which was - * Copyright (c) 2014,2016 Glenn Randers-Pehrson + * Derived from arm/filter_neon_intrinsics.c * - * Last changed in libpng 1.6.24 [August 4, 2016] + * Last changed in libpng 1.6.29 [March 16, 2017] * * This code is released under the libpng license. * For conditions of distribution and use, see the disclaimer * and license in png.h */ -#include "../../pngpriv.h" +#include "../pngpriv.h" #ifdef PNG_READ_SUPPORTED diff --git a/contrib/intel/intel_init.c b/intel/intel_init.c index 328e90e9a..8f08baf8c 100644 --- a/contrib/intel/intel_init.c +++ b/intel/intel_init.c @@ -1,19 +1,18 @@ /* intel_init.c - SSE2 optimized filter functions * - * Copyright (c) 2016 Google, Inc. - * Written by Mike Klein and Matt Sarett - * Derived from arm/arm_init.c, which was - * Copyright (c) 2014,2016 Glenn Randers-Pehrson + * Copyright (c) 2016-2017 Glenn Randers-Pehrson + * Written by Mike Klein and Matt Sarett, Google, Inc. + * Derived from arm/arm_init.c * - * Last changed in libpng 1.6.22 [May 26, 2016] + * Last changed in libpng 1.6.29 [March 16, 2017] * * This code is released under the libpng license. * For conditions of distribution and use, see the disclaimer * and license in png.h */ -#include "../../pngpriv.h" +#include "../pngpriv.h" #ifdef PNG_READ_SUPPORTED #if PNG_INTEL_SSE_IMPLEMENTATION > 0 diff --git a/libpng-manual.txt b/libpng-manual.txt index 048ba31c0..57ebbda4c 100644 --- a/libpng-manual.txt +++ b/libpng-manual.txt @@ -1,6 +1,6 @@ libpng-manual.txt - A description on how to use and modify libpng - libpng version 1.6.28 - January 5, 2017 + libpng version 1.6.29 - March 16, 2017 Updated and distributed by Glenn Randers-Pehrson <glennrp at users.sourceforge.net> Copyright (c) 1998-2016 Glenn Randers-Pehrson @@ -11,7 +11,7 @@ libpng-manual.txt - A description on how to use and modify libpng Based on: - libpng versions 0.97, January 1998, through 1.6.28 - January 5, 2017 + libpng versions 0.97, January 1998, through 1.6.29 - March 16, 2017 Updated and distributed by Glenn Randers-Pehrson Copyright (c) 1998-2016 Glenn Randers-Pehrson @@ -5355,7 +5355,7 @@ Since the PNG Development group is an ad-hoc body, we can't make an official declaration. This is your unofficial assurance that libpng from version 0.71 and -upward through 1.6.28 are Y2K compliant. It is my belief that earlier +upward through 1.6.29 are Y2K compliant. It is my belief that earlier versions were also Y2K compliant. Libpng only has two year fields. One is a 2-byte unsigned integer @@ -1,6 +1,6 @@ -.TH LIBPNG 3 "January 5, 2017" +.TH LIBPNG 3 "March 16, 2017" .SH NAME -libpng \- Portable Network Graphics (PNG) Reference Library 1.6.28 +libpng \- Portable Network Graphics (PNG) Reference Library 1.6.29 .SH SYNOPSIS \fB #include <png.h>\fP @@ -510,7 +510,7 @@ Following is a copy of the libpng-manual.txt file that accompanies libpng. .SH LIBPNG.TXT libpng-manual.txt - A description on how to use and modify libpng - libpng version 1.6.28 - January 5, 2017 + libpng version 1.6.29 - March 16, 2017 Updated and distributed by Glenn Randers-Pehrson <glennrp at users.sourceforge.net> Copyright (c) 1998-2016 Glenn Randers-Pehrson @@ -521,7 +521,7 @@ libpng-manual.txt - A description on how to use and modify libpng Based on: - libpng versions 0.97, January 1998, through 1.6.28 - January 5, 2017 + libpng versions 0.97, January 1998, through 1.6.29 - March 16, 2017 Updated and distributed by Glenn Randers-Pehrson Copyright (c) 1998-2016 Glenn Randers-Pehrson @@ -5865,7 +5865,7 @@ Since the PNG Development group is an ad-hoc body, we can't make an official declaration. This is your unofficial assurance that libpng from version 0.71 and -upward through 1.6.28 are Y2K compliant. It is my belief that earlier +upward through 1.6.29 are Y2K compliant. It is my belief that earlier versions were also Y2K compliant. Libpng only has two year fields. One is a 2-byte unsigned integer @@ -5967,7 +5967,7 @@ the first widely used release: ... 1.5.28 15 10528 15.so.15.28[.0] ... - 1.6.28 16 10628 16.so.16.28[.0] + 1.6.29 16 10629 16.so.16.29[.0] Henceforth the source version will match the shared-library minor and patch numbers; the shared-library major version number will be @@ -6023,7 +6023,7 @@ possible without all of you. Thanks to Frank J. T. Wojcik for helping with the documentation. -Libpng version 1.6.28 - January 5, 2017: +Libpng version 1.6.29 - March 16, 2017: Initially created in 1995 by Guy Eric Schalnat, then of Group 42, Inc. Currently maintained by Glenn Randers-Pehrson (glennrp at users.sourceforge.net). @@ -6048,7 +6048,7 @@ this sentence. This code is released under the libpng license. -libpng versions 1.0.7, July 1, 2000 through 1.6.28, January 5, 2017 are +libpng versions 1.0.7, July 1, 2000 through 1.6.29, March 16, 2017 are Copyright (c) 2000-2002, 2004, 2006-2017 Glenn Randers-Pehrson, are derived from libpng-1.0.6, and are distributed according to the same disclaimer and license as libpng-1.0.6 with the following individuals @@ -6061,6 +6061,8 @@ added to the list of Contributing Authors: Gilles Vollant James Yu Mandar Sahastrabuddhe + Google Inc. + Vadim Barkov and with the following additions to the disclaimer: @@ -6174,7 +6176,7 @@ files "pngbar.png" and "pngbar.jpg (88x31) and "pngnow.png" (98x31). Glenn Randers-Pehrson glennrp at users.sourceforge.net -January 5, 2017 +March 16, 2017 .\" end of man page diff --git a/libpngpf.3 b/libpngpf.3 index 585fd24c7..04e4b3b47 100644 --- a/libpngpf.3 +++ b/libpngpf.3 @@ -1,6 +1,6 @@ -.TH LIBPNGPF 3 "January 5, 2017" +.TH LIBPNGPF 3 "March 16, 2017" .SH NAME -libpng \- Portable Network Graphics (PNG) Reference Library 1.6.28 +libpng \- Portable Network Graphics (PNG) Reference Library 1.6.29 (private functions) .SH SYNOPSIS \fB#include \fI"pngpriv.h" @@ -1,4 +1,4 @@ -.TH PNG 5 "January 5, 2017" +.TH PNG 5 "March 16, 2017" .SH NAME png \- Portable Network Graphics (PNG) format .SH DESCRIPTION @@ -1,7 +1,7 @@ /* png.c - location for general purpose libpng functions * - * Last changed in libpng 1.6.28 [January 5, 2017] + * Last changed in libpng 1.6.29 [March 16, 2017] * Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) @@ -14,7 +14,7 @@ #include "pngpriv.h" /* Generate a compiler error if there is an old png.h in the search path. */ -typedef png_libpng_version_1_6_28 Your_png_h_is_not_version_1_6_28; +typedef png_libpng_version_1_6_29 Your_png_h_is_not_version_1_6_29; /* Tells libpng that we have already handled the first "num_bytes" bytes * of the PNG file signature. If the PNG data is embedded into another @@ -776,14 +776,14 @@ png_get_copyright(png_const_structrp png_ptr) #else # ifdef __STDC__ return PNG_STRING_NEWLINE \ - "libpng version 1.6.28 - January 5, 2017" PNG_STRING_NEWLINE \ + "libpng version 1.6.29 - March 16, 2017" PNG_STRING_NEWLINE \ "Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson" \ PNG_STRING_NEWLINE \ "Copyright (c) 1996-1997 Andreas Dilger" PNG_STRING_NEWLINE \ "Copyright (c) 1995-1996 Guy Eric Schalnat, Group 42, Inc." \ PNG_STRING_NEWLINE; # else - return "libpng version 1.6.28 - January 5, 2017\ + return "libpng version 1.6.29 - March 16, 2017\ Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson\ Copyright (c) 1996-1997 Andreas Dilger\ Copyright (c) 1995-1996 Guy Eric Schalnat, Group 42, Inc."; @@ -4260,13 +4260,13 @@ png_set_option(png_structrp png_ptr, int option, int onoff) if (png_ptr != NULL && option >= 0 && option < PNG_OPTION_NEXT && (option & 1) == 0) { - png_uint_32 mask = 3 << option; - png_uint_32 setting = (2 + (onoff != 0)) << option; + png_uint_32 mask = 3U << option; + png_uint_32 setting = (2U + (onoff != 0)) << option; png_uint_32 current = png_ptr->options; png_ptr->options = (png_uint_32)(((current & ~mask) | setting) & 0xff); - return (current & mask) >> option; + return (int)(current & mask) >> option; } return PNG_OPTION_INVALID; @@ -1,7 +1,7 @@ /* png.h - header file for PNG reference library * - * libpng version 1.6.28, January 5, 2017 + * libpng version 1.6.29, March 16, 2017 * * Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) @@ -12,7 +12,7 @@ * Authors and maintainers: * libpng versions 0.71, May 1995, through 0.88, January 1996: Guy Schalnat * libpng versions 0.89, June 1996, through 0.96, May 1997: Andreas Dilger - * libpng versions 0.97, January 1998, through 1.6.28, January 5, 2017: + * libpng versions 0.97, January 1998, through 1.6.29, March 16, 2017: * Glenn Randers-Pehrson. * See also "Contributing Authors", below. */ @@ -25,7 +25,7 @@ * * This code is released under the libpng license. * - * libpng versions 1.0.7, July 1, 2000 through 1.6.28, January 5, 2017 are + * libpng versions 1.0.7, July 1, 2000 through 1.6.29, March 16, 2017 are * Copyright (c) 2000-2002, 2004, 2006-2017 Glenn Randers-Pehrson, are * derived from libpng-1.0.6, and are distributed according to the same * disclaimer and license as libpng-1.0.6 with the following individuals @@ -38,6 +38,8 @@ * Gilles Vollant * James Yu * Mandar Sahastrabuddhe + * Google Inc. + * Vadim Barkov * * and with the following additions to the disclaimer: * @@ -211,7 +213,7 @@ * ... * 1.5.28 15 10527 15.so.15.28[.0] * ... - * 1.6.28 16 10628 16.so.16.28[.0] + * 1.6.29 16 10629 16.so.16.29[.0] * * Henceforth the source version will match the shared-library major * and minor numbers; the shared-library major version number will be @@ -239,13 +241,13 @@ * Y2K compliance in libpng: * ========================= * - * January 5, 2017 + * March 16, 2017 * * Since the PNG Development group is an ad-hoc body, we can't make * an official declaration. * * This is your unofficial assurance that libpng from version 0.71 and - * upward through 1.6.28 are Y2K compliant. It is my belief that + * upward through 1.6.29 are Y2K compliant. It is my belief that * earlier versions were also Y2K compliant. * * Libpng only has two year fields. One is a 2-byte unsigned integer @@ -307,8 +309,8 @@ */ /* Version information for png.h - this should match the version in png.c */ -#define PNG_LIBPNG_VER_STRING "1.6.28" -#define PNG_HEADER_VERSION_STRING " libpng version 1.6.28 - January 5, 2017\n" +#define PNG_LIBPNG_VER_STRING "1.6.29" +#define PNG_HEADER_VERSION_STRING " libpng version 1.6.29 - March 16, 2017\n" #define PNG_LIBPNG_VER_SONUM 16 #define PNG_LIBPNG_VER_DLLNUM 16 @@ -316,7 +318,7 @@ /* These should match the first 3 components of PNG_LIBPNG_VER_STRING: */ #define PNG_LIBPNG_VER_MAJOR 1 #define PNG_LIBPNG_VER_MINOR 6 -#define PNG_LIBPNG_VER_RELEASE 28 +#define PNG_LIBPNG_VER_RELEASE 29 /* This should match the numeric part of the final component of * PNG_LIBPNG_VER_STRING, omitting any leading zero: @@ -347,7 +349,7 @@ * version 1.0.0 was mis-numbered 100 instead of 10000). From * version 1.0.1 it's xxyyzz, where x=major, y=minor, z=release */ -#define PNG_LIBPNG_VER 10628 /* 1.6.28 */ +#define PNG_LIBPNG_VER 10629 /* 1.6.29 */ /* Library configuration: these options cannot be changed after * the library has been built. @@ -457,7 +459,7 @@ extern "C" { /* This triggers a compiler error in png.c, if png.c and png.h * do not agree upon the version number. */ -typedef char* png_libpng_version_1_6_28; +typedef char* png_libpng_version_1_6_29; /* Basic control structions. Read libpng-manual.txt or libpng.3 for more info. * @@ -3224,7 +3226,10 @@ PNG_EXPORT(245, int, png_image_write_to_memory, (png_imagep image, void *memory, # define PNG_MIPS_MSA 6 /* HARDWARE: MIPS Msa SIMD instructions supported */ #endif #define PNG_IGNORE_ADLER32 8 -#define PNG_OPTION_NEXT 10 /* Next option - numbers must be even */ +#ifdef PNG_POWERPC_VSX_API_SUPPORTED +# define PNG_POWERPC_VSX 10 /* HARDWARE: PowerPC VSX SIMD instructions supported */ +#endif +#define PNG_OPTION_NEXT 12 /* Next option - numbers must be even */ /* Return values: NOTE: there are four values and 'off' is *not* zero */ #define PNG_OPTION_UNSET 0 /* Unset - defaults to off */ @@ -1,7 +1,7 @@ /* pngconf.h - machine configurable file for libpng * - * libpng version 1.6.28, January 5, 2017 + * libpng version 1.6.29, March 16, 2017 * * Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) @@ -1,8 +1,8 @@ /* pngpriv.h - private declarations for use inside libpng * - * Last changed in libpng 1.6.26 [October 20, 2016] - * Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson + * Last changed in libpng 1.6.29 [March 16, 2017] + * Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) * @@ -190,6 +190,50 @@ # endif #endif +#ifndef PNG_POWERPC_VSX_OPT +# if defined(__PPC64__) && defined(__ALTIVEC__) && defined(__VSX__) +# define PNG_POWERPC_VSX_OPT 2 +# else +# define PNG_POWERPC_VSX_OPT 0 +# endif +#endif + +#ifndef PNG_INTEL_SSE_OPT +# ifdef PNG_INTEL_SSE + /* Only check for SSE if the build configuration has been modified to + * enable SSE optimizations. This means that these optimizations will + * be off by default. See contrib/intel for more details. + */ +# if defined(__SSE4_1__) || defined(__AVX__) || defined(__SSSE3__) || \ + defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \ + (defined(_M_IX86_FP) && _M_IX86_FP >= 2) +# define PNG_INTEL_SSE_OPT 1 +# endif +# endif +#endif + +#if PNG_INTEL_SSE_OPT > 0 +# ifndef PNG_INTEL_SSE_IMPLEMENTATION +# if defined(__SSE4_1__) || defined(__AVX__) + /* We are not actually using AVX, but checking for AVX is the best + way we can detect SSE4.1 and SSSE3 on MSVC. + */ +# define PNG_INTEL_SSE_IMPLEMENTATION 3 +# elif defined(__SSSE3__) +# define PNG_INTEL_SSE_IMPLEMENTATION 2 +# elif defined(__SSE2__) || defined(_M_X64) || defined(_M_AMD64) || \ + (defined(_M_IX86_FP) && _M_IX86_FP >= 2) +# define PNG_INTEL_SSE_IMPLEMENTATION 1 +# else +# define PNG_INTEL_SSE_IMPLEMENTATION 0 +# endif +# endif + +# if PNG_INTEL_SSE_IMPLEMENTATION > 0 +# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_sse2 +# endif +#endif + #if PNG_MIPS_MSA_OPT > 0 # define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_msa # ifndef PNG_MIPS_MSA_IMPLEMENTATION @@ -210,6 +254,11 @@ # endif #endif /* PNG_MIPS_MSA_OPT > 0 */ +#if PNG_POWERPC_VSX_OPT > 0 +# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_vsx +# define PNG_POWERPC_VSX_IMPLEMENTATION 1 +#endif + /* Is this a build of a DLL where compilation of the object modules requires * different preprocessor settings to those required for a simple library? If @@ -1256,6 +1305,38 @@ PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_msa,(png_row_infop row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); #endif +#if PNG_POWERPC_VSX_OPT > 0 +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_up_vsx,(png_row_infop row_info, + png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_vsx,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_vsx,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_vsx,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_vsx,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_vsx,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_vsx,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +#endif + +#if PNG_INTEL_SSE_IMPLEMENTATION > 0 +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_sse2,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_sse2,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_sse2,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_sse2,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_sse2,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_sse2,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +#endif + /* Choose the best filter to use and filter the row data */ PNG_INTERNAL_FUNCTION(void,png_write_find_filter,(png_structrp png_ptr, png_row_infop row_info),PNG_EMPTY); @@ -1991,6 +2072,11 @@ PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_neon, PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_msa, (png_structp png_ptr, unsigned int bpp), PNG_EMPTY); #endif + +# if PNG_INTEL_SSE_IMPLEMENTATION > 0 +PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_sse2, + (png_structp png_ptr, unsigned int bpp), PNG_EMPTY); +# endif #endif PNG_INTERNAL_FUNCTION(png_uint_32, png_check_keyword, (png_structrp png_ptr, diff --git a/pngrtran.c b/pngrtran.c index 0b4f4f906..16c1ed608 100644 --- a/pngrtran.c +++ b/pngrtran.c @@ -1,8 +1,8 @@ /* pngrtran.c - transforms the data in a row for PNG readers * - * Last changed in libpng 1.6.24 [August 4, 2016] - * Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson + * Last changed in libpng 1.6.29 [March 16, 2017] + * Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) * @@ -4302,7 +4302,7 @@ png_do_expand_palette(png_row_infop row_info, png_bytep row, if (num_trans > 0) { sp = row + (png_size_t)row_width - 1; - dp = row + (png_size_t)(row_width << 2) - 1; + dp = row + ((png_size_t)row_width << 2) - 1; for (i = 0; i < row_width; i++) { @@ -4463,7 +4463,7 @@ png_do_expand(png_row_infop row_info, png_bytep row, { gray = gray & 0xff; sp = row + (png_size_t)row_width - 1; - dp = row + (png_size_t)(row_width << 1) - 1; + dp = row + ((png_size_t)row_width << 1) - 1; for (i = 0; i < row_width; i++) { @@ -4519,7 +4519,7 @@ png_do_expand(png_row_infop row_info, png_bytep row, png_byte green = (png_byte)(trans_color->green & 0xff); png_byte blue = (png_byte)(trans_color->blue & 0xff); sp = row + (png_size_t)row_info->rowbytes - 1; - dp = row + (png_size_t)(row_width << 2) - 1; + dp = row + ((png_size_t)row_width << 2) - 1; for (i = 0; i < row_width; i++) { if (*(sp - 2) == red && *(sp - 1) == green && *(sp) == blue) @@ -4542,7 +4542,7 @@ png_do_expand(png_row_infop row_info, png_bytep row, png_byte green_low = (png_byte)(trans_color->green & 0xff); png_byte blue_low = (png_byte)(trans_color->blue & 0xff); sp = row + row_info->rowbytes - 1; - dp = row + (png_size_t)(row_width << 3) - 1; + dp = row + ((png_size_t)row_width << 3) - 1; for (i = 0; i < row_width; i++) { if (*(sp - 5) == red_high && diff --git a/pngrutil.c b/pngrutil.c index bee0ea115..76f079a69 100644 --- a/pngrutil.c +++ b/pngrutil.c @@ -1,8 +1,8 @@ /* pngrutil.c - utilities to read a PNG file * - * Last changed in libpng 1.6.27 [January 5, 2017] - * Copyright (c) 1998-2002,2004,2006-2016 Glenn Randers-Pehrson + * Last changed in libpng 1.6.29 [March 16, 2017] + * Copyright (c) 1998-2002,2004,2006-2017 Glenn Randers-Pehrson * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) * @@ -418,7 +418,7 @@ png_inflate_claim(png_structrp png_ptr, png_uint_32 owner) png_ptr->flags |= PNG_FLAG_ZSTREAM_INITIALIZED; } -#if ZLIB_VERNUM >= 0x1281 && \ +#if ZLIB_VERNUM >= 0x1290 && \ defined(PNG_SET_OPTION_SUPPORTED) && defined(PNG_IGNORE_ADLER32) if (((png_ptr->options >> PNG_IGNORE_ADLER32) & 3) == PNG_OPTION_ON) /* Turn off validation of the ADLER32 checksum in IDAT chunks */ @@ -2088,4 +2088,4 @@ main(void) #endif /* Generate a compiler error if there is an old png.h in the search path. */ -typedef png_libpng_version_1_6_28 Your_png_h_is_not_version_1_6_28; +typedef png_libpng_version_1_6_29 Your_png_h_is_not_version_1_6_29; diff --git a/pngwutil.c b/pngwutil.c index d1a82d45e..0f98d582d 100644 --- a/pngwutil.c +++ b/pngwutil.c @@ -675,6 +675,7 @@ png_write_IHDR(png_structrp png_ptr, png_uint_32 width, png_uint_32 height, int interlace_type) { png_byte buf[13]; /* Buffer to store the IHDR info */ + int is_invalid_depth; png_debug(1, "in png_write_IHDR"); @@ -700,11 +701,11 @@ png_write_IHDR(png_structrp png_ptr, png_uint_32 width, png_uint_32 height, break; case PNG_COLOR_TYPE_RGB: + is_invalid_depth = (bit_depth != 8); #ifdef PNG_WRITE_16BIT_SUPPORTED - if (bit_depth != 8 && bit_depth != 16) -#else - if (bit_depth != 8) + is_invalid_depth = (is_invalid_depth && bit_depth != 16); #endif + if (is_invalid_depth) png_error(png_ptr, "Invalid bit depth for RGB image"); png_ptr->channels = 3; @@ -726,18 +727,22 @@ png_write_IHDR(png_structrp png_ptr, png_uint_32 width, png_uint_32 height, break; case PNG_COLOR_TYPE_GRAY_ALPHA: - if (bit_depth != 8 && bit_depth != 16) + is_invalid_depth = (bit_depth != 8); +#ifdef PNG_WRITE_16BIT_SUPPORTED + is_invalid_depth = (is_invalid_depth && bit_depth != 16); +#endif + if (is_invalid_depth) png_error(png_ptr, "Invalid bit depth for grayscale+alpha image"); png_ptr->channels = 2; break; case PNG_COLOR_TYPE_RGB_ALPHA: + is_invalid_depth = (bit_depth != 8); #ifdef PNG_WRITE_16BIT_SUPPORTED - if (bit_depth != 8 && bit_depth != 16) -#else - if (bit_depth != 8) + is_invalid_depth = (is_invalid_depth && bit_depth != 16); #endif + if (is_invalid_depth) png_error(png_ptr, "Invalid bit depth for RGBA image"); png_ptr->channels = 4; diff --git a/powerpc/filter_vsx_intrinsics.c b/powerpc/filter_vsx_intrinsics.c new file mode 100644 index 000000000..e3de496bd --- /dev/null +++ b/powerpc/filter_vsx_intrinsics.c @@ -0,0 +1,767 @@ +/* filter_vsx_intrinsics.c - PowerPC optimised filter functions + * + * Copyright (c) 2017 Glenn Randers-Pehrson + * Written by Vadim Barkov, 2017. + * Last changed in libpng 1.6.29 [March 16, 2017] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ +#include <stdio.h> +#include <stdint.h> +#include "../pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +/* This code requires -maltivec and -mvsx on the command line: */ +#if PNG_POWERPC_VSX_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */ + +#include <altivec.h> + +#if PNG_POWERPC_VSX_OPT > 0 + +#ifndef __VSX__ +# error "This code requires VSX support (POWER7 and later). Please provide -mvsx compiler flag." +#endif + +#define vec_ld_unaligned(vec,data) vec = vec_vsx_ld(0,data) +#define vec_st_unaligned(vec,data) vec_vsx_st(vec,0,data) + + +/* Functions in this file look at most 3 pixels (a,b,c) to predict the 4th (d). + * They're positioned like this: + * prev: c b + * row: a d + * The Sub filter predicts d=a, Avg d=(a+b)/2, and Paeth predicts d to be + * whichever of a, b, or c is closest to p=a+b-c. + * ( this is taken from ../intel/filter_sse2_intrinsics.c ) + */ + +#define vsx_declare_common_vars(row_info,row,prev_row,offset) \ + png_byte i;\ + png_bytep rp = row + offset;\ + png_const_bytep pp = prev_row;\ + png_size_t unaligned_top = 16 - (((png_size_t)rp % 16));\ + png_size_t istop;\ + if(unaligned_top == 16)\ + unaligned_top = 0;\ + istop = row_info->rowbytes;\ + if((unaligned_top < istop))\ + istop -= unaligned_top;\ + else{\ + unaligned_top = istop;\ + istop = 0;\ + } + +void png_read_filter_row_up_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + vector unsigned char rp_vec; + vector unsigned char pp_vec; + vsx_declare_common_vars(row_info,row,prev_row,0) + + /* Altivec operations require 16-byte aligned data + * but input can be unaligned. So we calculate + * unaligned part as usual. + */ + for (i = 0; i < unaligned_top; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff); + rp++; + } + + /* Using SIMD while we can */ + while( istop >= 16 ) + { + rp_vec = vec_ld(0,rp); + vec_ld_unaligned(pp_vec,pp); + + rp_vec = vec_add(rp_vec,pp_vec); + + vec_st(rp_vec,0,rp); + + pp += 16; + rp += 16; + istop -= 16; + } + + if(istop > 0) + { + /* If byte count of row is not divisible by 16 + * we will process remaining part as usual + */ + for (i = 0; i < istop; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff); + rp++; + } +} + +} + +static const vector unsigned char VSX_LEFTSHIFTED1_4 = {16,16,16,16, 0, 1, 2, 3,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_LEFTSHIFTED2_4 = {16,16,16,16,16,16,16,16, 4, 5, 6, 7,16,16,16,16}; +static const vector unsigned char VSX_LEFTSHIFTED3_4 = {16,16,16,16,16,16,16,16,16,16,16,16, 8, 9,10,11}; + +static const vector unsigned char VSX_LEFTSHIFTED1_3 = {16,16,16, 0, 1, 2,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_LEFTSHIFTED2_3 = {16,16,16,16,16,16, 3, 4, 5,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_LEFTSHIFTED3_3 = {16,16,16,16,16,16,16,16,16, 6, 7, 8,16,16,16,16}; +static const vector unsigned char VSX_LEFTSHIFTED4_3 = {16,16,16,16,16,16,16,16,16,16,16,16, 9,10,11,16}; + +static const vector unsigned char VSX_NOT_SHIFTED1_4 = {16,16,16,16, 4, 5, 6, 7,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_NOT_SHIFTED2_4 = {16,16,16,16,16,16,16,16, 8, 9,10,11,16,16,16,16}; +static const vector unsigned char VSX_NOT_SHIFTED3_4 = {16,16,16,16,16,16,16,16,16,16,16,16,12,13,14,15}; + +static const vector unsigned char VSX_NOT_SHIFTED1_3 = {16,16,16, 3, 4, 5,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_NOT_SHIFTED2_3 = {16,16,16,16,16,16, 6, 7, 8,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_NOT_SHIFTED3_3 = {16,16,16,16,16,16,16,16,16, 9,10,11,16,16,16,16}; +static const vector unsigned char VSX_NOT_SHIFTED4_3 = {16,16,16,16,16,16,16,16,16,16,16,16,12,13,14,16}; + +static const vector unsigned char VSX_CHAR_ZERO = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}; +#ifdef __LITTLE_ENDIAN__ + +static const vector unsigned char VSX_CHAR_TO_SHORT1_4 = { 4,16, 5,16, 6,16, 7,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT2_4 = { 8,16, 9,16,10,16,11,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT3_4 = {12,16,13,16,14,16,15,16,16,16,16,16,16,16,16,16}; + +static const vector unsigned char VSX_SHORT_TO_CHAR1_4 = {16,16,16,16, 0, 2, 4, 6,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR2_4 = {16,16,16,16,16,16,16,16, 0, 2, 4, 6,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR3_4 = {16,16,16,16,16,16,16,16,16,16,16,16, 0, 2, 4, 6}; + +static const vector unsigned char VSX_CHAR_TO_SHORT1_3 = { 3,16, 4,16, 5,16,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT2_3 = { 6,16, 7,16, 8,16,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT3_3 = { 9,16,10,16,11,16,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT4_3 = {12,16,13,16,14,16,16,16,16,16,16,16,16,16,16,16}; + +static const vector unsigned char VSX_SHORT_TO_CHAR1_3 = {16,16,16, 0, 2, 4,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR2_3 = {16,16,16,16,16,16, 0, 2, 4,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR3_3 = {16,16,16,16,16,16,16,16,16, 0, 2, 4,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR4_3 = {16,16,16,16,16,16,16,16,16,16,16,16, 0, 2, 4,16}; + +#elif defined(__BIG_ENDIAN__) + +static const vector unsigned char VSX_CHAR_TO_SHORT1_4 = {16, 4,16, 5,16, 6,16, 7,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT2_4 = {16, 8,16, 9,16,10,16,11,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT3_4 = {16,12,16,13,16,14,16,15,16,16,16,16,16,16,16,16}; + +static const vector unsigned char VSX_SHORT_TO_CHAR1_4 = {16,16,16,16, 1, 3, 5, 7,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR2_4 = {16,16,16,16,16,16,16,16, 1, 3, 5, 7,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR3_4 = {16,16,16,16,16,16,16,16,16,16,16,16, 1, 3, 5, 7}; + +static const vector unsigned char VSX_CHAR_TO_SHORT1_3 = {16, 3,16, 4,16, 5,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT2_3 = {16, 6,16, 7,16, 8,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT3_3 = {16, 9,16,10,16,11,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_CHAR_TO_SHORT4_3 = {16,12,16,13,16,14,16,16,16,16,16,16,16,16,16,16}; + +static const vector unsigned char VSX_SHORT_TO_CHAR1_3 = {16,16,16, 1, 3, 5,16,16,16,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR2_3 = {16,16,16,16,16,16, 1, 3, 5,16,16,16,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR3_3 = {16,16,16,16,16,16,16,16,16, 1, 3, 5,16,16,16,16}; +static const vector unsigned char VSX_SHORT_TO_CHAR4_3 = {16,16,16,16,16,16,16,16,16,16,16,16, 1, 3, 5,16}; + +#endif + +#define vsx_char_to_short(vec,offset,bpp) (vector unsigned short)vec_perm((vec),VSX_CHAR_ZERO,VSX_CHAR_TO_SHORT##offset##_##bpp) +#define vsx_short_to_char(vec,offset,bpp) vec_perm(((vector unsigned char)(vec)),VSX_CHAR_ZERO,VSX_SHORT_TO_CHAR##offset##_##bpp) + +#ifdef PNG_USE_ABS +# define vsx_abs(number) abs(number) +#else +# define vsx_abs(number) (number > 0) ? (number) : -(number) +#endif + +void png_read_filter_row_sub4_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + const png_byte bpp = 4; + + vector unsigned char rp_vec; + vector unsigned char part_vec; + + vsx_declare_common_vars(row_info,row,prev_row,bpp) + + PNG_UNUSED(pp) + + /* Altivec operations require 16-byte aligned data + * but input can be unaligned. So we calculate + * unaligned part as usual. + */ + for (i = 0; i < unaligned_top; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } + + /* Using SIMD while we can */ + while( istop >= 16 ) + { + for(i=0;i < bpp ; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } + rp -= bpp; + + rp_vec = vec_ld(0,rp); + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_4); + rp_vec = vec_add(rp_vec,part_vec); + + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_4); + rp_vec = vec_add(rp_vec,part_vec); + + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_4); + rp_vec = vec_add(rp_vec,part_vec); + + vec_st(rp_vec,0,rp); + + rp += 16; + istop -= 16; + } + + if(istop > 0) + for (i = 0; i < istop % 16; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp - bpp))) & 0xff); + rp++; + } + +} + +void png_read_filter_row_sub3_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + const png_byte bpp = 3; + + vector unsigned char rp_vec; + vector unsigned char part_vec; + + vsx_declare_common_vars(row_info,row,prev_row,bpp) + + PNG_UNUSED(pp) + + /* Altivec operations require 16-byte aligned data + * but input can be unaligned. So we calculate + * unaligned part as usual. + */ + for (i = 0; i < unaligned_top; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } + + /* Using SIMD while we can */ + while( istop >= 16 ) + { + for(i=0;i < bpp ; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } + rp -= bpp; + + rp_vec = vec_ld(0,rp); + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_3); + rp_vec = vec_add(rp_vec,part_vec); + + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_3); + rp_vec = vec_add(rp_vec,part_vec); + + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_3); + rp_vec = vec_add(rp_vec,part_vec); + + part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED4_3); + rp_vec = vec_add(rp_vec,part_vec); + + vec_st(rp_vec,0,rp); + rp += 15; + istop -= 16; + + /* Since 16 % bpp = 16 % 3 = 1, last element of array must + * be proceeded manually + */ + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } + + if(istop > 0) + for (i = 0; i < istop % 16; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } +} + +void png_read_filter_row_avg4_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + const png_byte bpp = 4; + + vector unsigned char rp_vec; + vector unsigned char pp_vec; + vector unsigned char pp_part_vec; + vector unsigned char rp_part_vec; + vector unsigned char avg_vec; + + vsx_declare_common_vars(row_info,row,prev_row,bpp) + rp -= bpp; + if(istop >= bpp) + istop -= bpp; + + for (i = 0; i < bpp; i++) + { + *rp = (png_byte)(((int)(*rp) + + ((int)(*pp++) / 2 )) & 0xff); + + rp++; + } + + /* Altivec operations require 16-byte aligned data + * but input can be unaligned. So we calculate + * unaligned part as usual. + */ + for (i = 0; i < unaligned_top; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } + + /* Using SIMD while we can */ + while( istop >= 16 ) + { + for(i=0;i < bpp ; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } + rp -= bpp; + pp -= bpp; + + vec_ld_unaligned(pp_vec,pp); + rp_vec = vec_ld(0,rp); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_4); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED1_4); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_4); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED2_4); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_4); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED3_4); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + vec_st(rp_vec,0,rp); + + rp += 16; + pp += 16; + istop -= 16; + } + + if(istop > 0) + for (i = 0; i < istop % 16; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } +} + +void png_read_filter_row_avg3_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + const png_byte bpp = 3; + + vector unsigned char rp_vec; + vector unsigned char pp_vec; + vector unsigned char pp_part_vec; + vector unsigned char rp_part_vec; + vector unsigned char avg_vec; + + vsx_declare_common_vars(row_info,row,prev_row,bpp) + rp -= bpp; + if(istop >= bpp) + istop -= bpp; + + for (i = 0; i < bpp; i++) + { + *rp = (png_byte)(((int)(*rp) + + ((int)(*pp++) / 2 )) & 0xff); + + rp++; + } + + /* Altivec operations require 16-byte aligned data + * but input can be unaligned. So we calculate + * unaligned part as usual. + */ + for (i = 0; i < unaligned_top; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } + + /* Using SIMD while we can */ + while( istop >= 16 ) + { + for(i=0;i < bpp ; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } + rp -= bpp; + pp -= bpp; + + vec_ld_unaligned(pp_vec,pp); + rp_vec = vec_ld(0,rp); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED1_3); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED1_3); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED2_3); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED2_3); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED3_3); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED3_3); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + rp_part_vec = vec_perm(rp_vec,VSX_CHAR_ZERO,VSX_LEFTSHIFTED4_3); + pp_part_vec = vec_perm(pp_vec,VSX_CHAR_ZERO,VSX_NOT_SHIFTED4_3); + avg_vec = vec_avg(rp_part_vec,pp_part_vec); + avg_vec = vec_sub(avg_vec, vec_and(vec_xor(rp_part_vec,pp_part_vec),vec_splat_u8(1))); + rp_vec = vec_add(rp_vec,avg_vec); + + vec_st(rp_vec,0,rp); + + rp += 15; + pp += 15; + istop -= 16; + + /* Since 16 % bpp = 16 % 3 = 1, last element of array must + * be proceeded manually + */ + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + rp++; + } + + if(istop > 0) + for (i = 0; i < istop % 16; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } +} + +/* Bytewise c ? t : e. */ +#define if_then_else(c,t,e) vec_sel(e,t,c) + +#define vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) {\ + c = *(pp - bpp);\ + a = *(rp - bpp);\ + b = *pp++;\ + p = b - c;\ + pc = a - c;\ + pa = vsx_abs(p);\ + pb = vsx_abs(pc);\ + pc = vsx_abs(p + pc);\ + if (pb < pa) pa = pb, a = b;\ + if (pc < pa) a = c;\ + a += *rp;\ + *rp++ = (png_byte)a;\ + } + +void png_read_filter_row_paeth4_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + const png_byte bpp = 4; + + int a, b, c, pa, pb, pc, p; + vector unsigned char rp_vec; + vector unsigned char pp_vec; + vector unsigned short a_vec,b_vec,c_vec,nearest_vec; + vector signed short pa_vec,pb_vec,pc_vec,smallest_vec; + + vsx_declare_common_vars(row_info,row,prev_row,bpp) + rp -= bpp; + if(istop >= bpp) + istop -= bpp; + + /* Process the first pixel in the row completely (this is the same as 'up' + * because there is only one candidate predictor for the first row). + */ + for(i = 0; i < bpp ; i++) + { + *rp = (png_byte)( *rp + *pp); + rp++; + pp++; + } + + for(i = 0; i < unaligned_top ; i++) + { + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } + + while( istop >= 16) + { + for(i = 0; i < bpp ; i++) + { + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } + + rp -= bpp; + pp -= bpp; + rp_vec = vec_ld(0,rp); + vec_ld_unaligned(pp_vec,pp); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_4),1,4); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED1_4),1,4); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_4),1,4); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,1,4))); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_4),2,4); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED2_4),2,4); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_4),2,4); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,2,4))); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_4),3,4); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED3_4),3,4); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_4),3,4); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,3,4))); + + vec_st(rp_vec,0,rp); + + rp += 16; + pp += 16; + istop -= 16; + } + + if(istop > 0) + for (i = 0; i < istop % 16; i++) + { + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } +} + +void png_read_filter_row_paeth3_vsx(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + const png_byte bpp = 3; + + int a, b, c, pa, pb, pc, p; + vector unsigned char rp_vec; + vector unsigned char pp_vec; + vector unsigned short a_vec,b_vec,c_vec,nearest_vec; + vector signed short pa_vec,pb_vec,pc_vec,smallest_vec; + + vsx_declare_common_vars(row_info,row,prev_row,bpp) + rp -= bpp; + if(istop >= bpp) + istop -= bpp; + + /* Process the first pixel in the row completely (this is the same as 'up' + * because there is only one candidate predictor for the first row). + */ + for(i = 0; i < bpp ; i++) + { + *rp = (png_byte)( *rp + *pp); + rp++; + pp++; + } + + for(i = 0; i < unaligned_top ; i++) + { + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } + + while( istop >= 16) + { + for(i = 0; i < bpp ; i++) + { + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } + + rp -= bpp; + pp -= bpp; + rp_vec = vec_ld(0,rp); + vec_ld_unaligned(pp_vec,pp); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_3),1,3); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED1_3),1,3); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED1_3),1,3); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,1,3))); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_3),2,3); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED2_3),2,3); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED2_3),2,3); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,2,3))); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_3),3,3); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED3_3),3,3); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED3_3),3,3); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,3,3))); + + a_vec = vsx_char_to_short(vec_perm(rp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED4_3),4,3); + b_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_NOT_SHIFTED4_3),4,3); + c_vec = vsx_char_to_short(vec_perm(pp_vec , VSX_CHAR_ZERO , VSX_LEFTSHIFTED4_3),4,3); + pa_vec = (vector signed short) vec_sub(b_vec,c_vec); + pb_vec = (vector signed short) vec_sub(a_vec , c_vec); + pc_vec = vec_add(pa_vec,pb_vec); + pa_vec = vec_abs(pa_vec); + pb_vec = vec_abs(pb_vec); + pc_vec = vec_abs(pc_vec); + smallest_vec = vec_min(pc_vec, vec_min(pa_vec,pb_vec)); + nearest_vec = if_then_else( + vec_cmpeq(pa_vec,smallest_vec), + a_vec, + if_then_else( + vec_cmpeq(pb_vec,smallest_vec), + b_vec, + c_vec + ) + ); + rp_vec = vec_add(rp_vec,(vsx_short_to_char(nearest_vec,4,3))); + + vec_st(rp_vec,0,rp); + + rp += 15; + pp += 15; + istop -= 16; + + /* Since 16 % bpp = 16 % 3 = 1, last element of array must + * be proceeded manually + */ + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } + + if(istop > 0) + for (i = 0; i < istop % 16; i++) + { + vsx_paeth_process(rp,pp,a,b,c,pa,pb,pc,bpp) + } +} + +#endif /* PNG_POWERPC_VSX_OPT > 0 */ +#endif /* PNG_POWERPC_VSX_IMPLEMENTATION == 1 (intrinsics) */ +#endif /* READ */ diff --git a/powerpc/powerpc_init.c b/powerpc/powerpc_init.c new file mode 100644 index 000000000..07016177c --- /dev/null +++ b/powerpc/powerpc_init.c @@ -0,0 +1,125 @@ + +/* powerpc_init.c - POWERPC optimised filter functions + * + * Copyright (c) 2017 Glenn Randers-Pehrson + * Written by Vadim Barkov, 2017. + * Last changed in libpng 1.6.29 [March 16, 2017] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ +/* Below, after checking __linux__, various non-C90 POSIX 1003.1 functions are + * called. + */ +#define _POSIX_SOURCE 1 + +#include <stdio.h> +#include "../pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +#if PNG_POWERPC_VSX_OPT > 0 +#ifdef PNG_POWERPC_VSX_CHECK_SUPPORTED /* Do run-time checks */ +/* WARNING: it is strongly recommended that you do not build libpng with + * run-time checks for CPU features if at all possible. In the case of the PowerPC + * VSX instructions there is no processor-specific way of detecting the + * presence of the required support, therefore run-time detection is extremely + * OS specific. + * + * You may set the macro PNG_POWERPC_VSX_FILE to the file name of file containing + * a fragment of C source code which defines the png_have_vsx function. There + * are a number of implementations in contrib/powerpc-vsx, but the only one that + * has partial support is contrib/powerpc-vsx/linux.c - a generic Linux + * implementation which reads /proc/cpufino. + */ +#ifndef PNG_POWERPC_VSX_FILE +# ifdef __linux__ +# define PNG_POWERPC_VSX_FILE "contrib/powerpc-vsx/linux_aux.c" +# endif +#endif + +#ifdef PNG_POWERPC_VSX_FILE + +#include <signal.h> /* for sig_atomic_t */ +static int png_have_vsx(png_structp png_ptr); +#include PNG_POWERPC_VSX_FILE + +#else /* PNG_POWERPC_VSX_FILE */ +# error "PNG_POWERPC_VSX_FILE undefined: no support for run-time POWERPC VSX checks" +#endif /* PNG_POWERPC_VSX_FILE */ +#endif /* PNG_POWERPC_VSX_CHECK_SUPPORTED */ + +void +png_init_filter_functions_vsx(png_structp pp, unsigned int bpp) +{ + /* The switch statement is compiled in for POWERPC_VSX_API, the call to + * png_have_vsx is compiled in for POWERPC_VSX_CHECK. If both are defined + * the check is only performed if the API has not set the PowerPC option on + * or off explicitly. In this case the check controls what happens. + */ + +#ifdef PNG_POWERPC_VSX_API_SUPPORTED + switch ((pp->options >> PNG_POWERPC_VSX) & 3) + { + case PNG_OPTION_UNSET: + /* Allow the run-time check to execute if it has been enabled - + * thus both API and CHECK can be turned on. If it isn't supported + * this case will fall through to the 'default' below, which just + * returns. + */ +#endif /* PNG_POWERPC_VSX_API_SUPPORTED */ +#ifdef PNG_POWERPC_VSX_CHECK_SUPPORTED + { + static volatile sig_atomic_t no_vsx = -1; /* not checked */ + + if (no_vsx < 0) + no_vsx = !png_have_vsx(pp); + + if (no_vsx) + return; + } +#ifdef PNG_POWERPC_VSX_API_SUPPORTED + break; +#endif +#endif /* PNG_POWERPC_VSX_CHECK_SUPPORTED */ + +#ifdef PNG_POWERPC_VSX_API_SUPPORTED + default: /* OFF or INVALID */ + return; + + case PNG_OPTION_ON: + /* Option turned on */ + break; + } +#endif + + /* IMPORTANT: any new internal functions used here must be declared using + * PNG_INTERNAL_FUNCTION in ../pngpriv.h. This is required so that the + * 'prefix' option to configure works: + * + * ./configure --with-libpng-prefix=foobar_ + * + * Verify you have got this right by running the above command, doing a build + * and examining pngprefix.h; it must contain a #define for every external + * function you add. (Notice that this happens automatically for the + * initialization function.) + */ + pp->read_filter[PNG_FILTER_VALUE_UP-1] = png_read_filter_row_up_vsx; + + if (bpp == 3) + { + pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub3_vsx; + pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg3_vsx; + pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth3_vsx; + } + + else if (bpp == 4) + { + pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub4_vsx; + pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg4_vsx; + pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = png_read_filter_row_paeth4_vsx; + } +} +#endif /* PNG_POWERPC_VSX_OPT > 0 */ +#endif /* READ */ diff --git a/projects/vstudio/README.txt b/projects/vstudio/README.txt index 9fd0f1dfb..1000d118e 100644 --- a/projects/vstudio/README.txt +++ b/projects/vstudio/README.txt @@ -1,7 +1,7 @@ VisualStudio instructions -libpng version 1.6.28 - January 5, 2017 +libpng version 1.6.29 - March 16, 2017 Copyright (c) 2010,2013,2015 Glenn Randers-Pehrson diff --git a/projects/vstudio/zlib.props b/projects/vstudio/zlib.props index e0e0d8fd6..18da337e4 100644 --- a/projects/vstudio/zlib.props +++ b/projects/vstudio/zlib.props @@ -2,7 +2,7 @@ <!-- * zlib.props - location of zlib source * - * libpng version 1.6.28 - January 5, 2017 + * libpng version 1.6.29 - March 16, 2017 * * Copyright (c) 1998-2011 Glenn Randers-Pehrson * diff --git a/scripts/README.txt b/scripts/README.txt index 1fbdb2e52..9b3c0907f 100644 --- a/scripts/README.txt +++ b/scripts/README.txt @@ -1,9 +1,9 @@ -Makefiles for libpng version 1.6.28 - January 5, 2017 +Makefiles for libpng version 1.6.29 - March 16, 2017 pnglibconf.h.prebuilt => Stores configuration settings makefile.linux => Linux/ELF makefile - (gcc, creates libpng16.so.16.1.6.28) + (gcc, creates libpng16.so.16.1.6.29) makefile.gcc => Generic makefile (gcc, creates static libpng.a) makefile.knr => Archaic UNIX Makefile that converts files with ansi2knr (Requires ansi2knr.c from @@ -33,12 +33,12 @@ pnglibconf.h.prebuilt => Stores configuration settings makefile.os2 => OS/2 Makefile (gcc and emx, requires libpng.def) makefile.sco => For SCO OSr5 ELF and Unixware 7 with Native cc makefile.sggcc => Silicon Graphics (gcc, - creates libpng16.so.16.1.6.28) + creates libpng16.so.16.1.6.29) makefile.sgi => Silicon Graphics IRIX makefile (cc, creates static lib) makefile.solaris => Solaris 2.X makefile (gcc, - creates libpng16.so.16.1.6.28) + creates libpng16.so.16.1.6.29) makefile.so9 => Solaris 9 makefile (gcc, - creates libpng16.so.16.1.6.28) + creates libpng16.so.16.1.6.29) makefile.std => Generic UNIX makefile (cc, creates static libpng.a) makefile.sunos => Sun makefile makefile.32sunu => Sun Ultra 32-bit makefile diff --git a/scripts/def.c b/scripts/def.c index 99396123b..6a1831bb5 100644 --- a/scripts/def.c +++ b/scripts/def.c @@ -21,7 +21,7 @@ PNG_DFN "OS2 DESCRIPTION "PNG image compression library"" PNG_DFN "OS2 CODE PRELOAD MOVEABLE DISCARDABLE" PNG_DFN "" PNG_DFN "EXPORTS" -PNG_DFN ";Version 1.6.28" +PNG_DFN ";Version 1.6.29" #define PNG_EXPORTA(ordinal, type, name, args, attributes)\ PNG_DFN "@" SYMBOL_PREFIX "@@" name "@" diff --git a/scripts/libpng-config-head.in b/scripts/libpng-config-head.in index 9b6b32bf4..e06ed8e12 100644 --- a/scripts/libpng-config-head.in +++ b/scripts/libpng-config-head.in @@ -11,7 +11,7 @@ # Modeled after libxml-config. -version=1.6.28 +version=1.6.29 prefix="" libdir="" libs="" diff --git a/scripts/libpng.pc.in b/scripts/libpng.pc.in index f363b3806..e8bf22dbc 100644 --- a/scripts/libpng.pc.in +++ b/scripts/libpng.pc.in @@ -5,6 +5,6 @@ includedir=@includedir@/libpng16 Name: libpng Description: Loads and saves PNG files -Version: 1.6.28 +Version: 1.6.29 Libs: -L${libdir} -lpng16 Cflags: -I${includedir} diff --git a/scripts/makefile.cegcc b/scripts/makefile.cegcc index 0fe9eac4b..ca8ef6aed 100644 --- a/scripts/makefile.cegcc +++ b/scripts/makefile.cegcc @@ -23,7 +23,7 @@ VERMAJ = 1 VERMIN = 6 -VERMIC = 28 +VERMIC = 29 VER = $(VERMAJ).$(VERMIN).$(VERMIC) NAME = libpng PACKAGE = $(NAME)-$(VER) diff --git a/scripts/makefile.linux b/scripts/makefile.linux index ccb866a05..c6450bb55 100644 --- a/scripts/makefile.linux +++ b/scripts/makefile.linux @@ -10,7 +10,7 @@ # Library name: LIBNAME = libpng16 PNGMAJ = 16 -RELEASE = 28 +RELEASE = 29 # Shared library names: LIBSO=$(LIBNAME).so diff --git a/scripts/makefile.msys b/scripts/makefile.msys index 72c31c158..4df88f915 100644 --- a/scripts/makefile.msys +++ b/scripts/makefile.msys @@ -18,7 +18,7 @@ exec_prefix=$(prefix) # Library name: LIBNAME = libpng16 PNGMAJ = 16 -RELEASE = 28 +RELEASE = 29 # Shared library names: LIBSO=$(LIBNAME).dll diff --git a/scripts/makefile.ne12bsd b/scripts/makefile.ne12bsd index 4fddc77f4..fd4e48550 100644 --- a/scripts/makefile.ne12bsd +++ b/scripts/makefile.ne12bsd @@ -17,7 +17,7 @@ INCSDIR=${LOCALBASE}/include/libpng16 LIB= png16 SHLIB_MAJOR= 0 -SHLIB_MINOR= 1.6.28 +SHLIB_MINOR= 1.6.29 SRCS= png.c pngset.c pngget.c pngrutil.c pngtrans.c pngwutil.c \ pngread.c pngrio.c pngwio.c pngwrite.c pngrtran.c \ pngwtran.c pngmem.c pngerror.c pngpread.c diff --git a/scripts/makefile.netbsd b/scripts/makefile.netbsd index 4688b62aa..210c8b606 100644 --- a/scripts/makefile.netbsd +++ b/scripts/makefile.netbsd @@ -17,7 +17,7 @@ INCSDIR=${LOCALBASE}/include LIB= png SHLIB_MAJOR= 16 -SHLIB_MINOR= 1.6.28 +SHLIB_MINOR= 1.6.29 SRCS= png.c pngset.c pngget.c pngrutil.c pngtrans.c pngwutil.c \ pngread.c pngrio.c pngwio.c pngwrite.c pngrtran.c \ pngwtran.c pngmem.c pngerror.c pngpread.c diff --git a/scripts/makefile.openbsd b/scripts/makefile.openbsd index ba9def35f..4f3d5823a 100644 --- a/scripts/makefile.openbsd +++ b/scripts/makefile.openbsd @@ -11,7 +11,7 @@ LIBDIR= ${PREFIX}/lib MANDIR= ${PREFIX}/man/cat SHLIB_MAJOR= 16 -SHLIB_MINOR= 1.6.28 +SHLIB_MINOR= 1.6.29 LIB= png SRCS= png.c pngerror.c pngget.c pngmem.c pngpread.c \ diff --git a/scripts/pnglibconf.dfa b/scripts/pnglibconf.dfa index 019c06d47..9df2a4311 100644 --- a/scripts/pnglibconf.dfa +++ b/scripts/pnglibconf.dfa @@ -229,6 +229,33 @@ option ARM_NEON_API disabled requires ALIGNED_MEMORY enables SET_OPTION, option ARM_NEON_CHECK disabled requires ALIGNED_MEMORY, sets ARM_NEON_OPT 1 +# These options are specific to the PowerPC VSX hardware optimizations. +# +# POWERPC_VSX_OPT: unset: check at compile time (__PPC64__,__ALTIVEC__,__VSX__ +# must be defined by the compiler, typically as a result +# of specifying +# "-mvsx -maltivec" compiler flags) +# 0: disable (even if the CPU supports VSX.) +# 1: check at run time (via POWERPC_VSX_{API,CHECK}) +# 2: switch on unconditionally (inadvisable - instead pass +# -mvsx -maltivec to compiler options) +# When building libpng avoid using any setting other than '0'; '1' is +# set automatically when either 'API' or 'CHECK' are configured in, +# '2' should not be necessary as "-mvsx -maltivec" will achieve the same +# effect as well as applying VSX optimizations to the rest of the +# libpng code. +# POWERPC_VSX_API: (PNG_POWERPC_VSX == 1) allow the optimization to be switched on +# with png_set_option +# POWERPC_VSX_CHECK: (PNG_POWERPC_VSX == 1) compile a run-time check to see if VSX +# extensions are supported. This is supported not for all OSes +# (see contrib/powerpc/README) +setting POWERPC_VSX_OPT +option POWERPC_VSX_API disabled enables SET_OPTION, + sets POWERPC_VSX_OPT 1 +option POWERPC_VSX_CHECK disabled, + sets POWERPC_VSX_OPT 1 + + # These settings configure the default compression level (0-9) and 'strategy'; # strategy is as defined by the implementors of zlib. It describes the input # data and modifies the zlib parameters in an attempt to optimize the balance diff --git a/scripts/pnglibconf.h.prebuilt b/scripts/pnglibconf.h.prebuilt index ee7057360..82f6ba461 100644 --- a/scripts/pnglibconf.h.prebuilt +++ b/scripts/pnglibconf.h.prebuilt @@ -1,8 +1,8 @@ -/* libpng 1.6.28 STANDARD API DEFINITION */ +/* libpng 1.6.29 STANDARD API DEFINITION */ /* pnglibconf.h - library build configuration */ -/* Libpng version 1.6.28 - January 5, 2017 */ +/* Libpng version 1.6.29 - March 16, 2017 */ /* Copyright (c) 1998-2015 Glenn Randers-Pehrson */ @@ -20,6 +20,8 @@ #define PNG_ALIGNED_MEMORY_SUPPORTED /*#undef PNG_ARM_NEON_API_SUPPORTED*/ /*#undef PNG_ARM_NEON_CHECK_SUPPORTED*/ +/*#undef PNG_POWERPC_VSX_API_SUPPORTED*/ +/*#undef PNG_POWERPC_VSX_CHECK_SUPPORTED*/ #define PNG_BENIGN_ERRORS_SUPPORTED #define PNG_BENIGN_READ_ERRORS_SUPPORTED /*#undef PNG_BENIGN_WRITE_ERRORS_SUPPORTED*/ diff --git a/scripts/symbols.def b/scripts/symbols.def index 4ebe7d0da..6c8eebff1 100644 --- a/scripts/symbols.def +++ b/scripts/symbols.def @@ -1,4 +1,4 @@ -;Version 1.6.28 +;Version 1.6.29 ;-------------------------------------------------------------- ; LIBPNG symbol list as a Win32 DEF file ; Contains all the symbols that can be exported from libpng |